Enhanced Detection of Alzheimer's Disease Using Weighted Ensemble Learning

Alzheimer's Disease (AD) is a progressive, irrevocable neurodegenerative disease that abolishes memory and other important mental functions: deteriorating the ability to communicate; causing difficulties in speaking or walking. Alzheimer's disease is most crucial health issue worldwide, in which early and precise diagnosis is important for interference and treatment. Recent developments have especially shown success in analysis of medical image, particularly in a field of AD detection using deep learning models. However, traditional deep learning methods are limited to exploit the complexity and heterogeneity of AD imaging data. This paper aims to fine-tune and validate the Deep Convolutional Networks for Alzheimer's disease diagnosis. This is a free accessible-source dataset comes from Kaggle and contains five classes by the different phases of Alzheimer's disease. In this paper, an attempt to alleviate the problems mentioned in this investigation by combining deep transfer learning models advanced pre-processing methods and ensemble strategies (weighted average) has shown significant improvements with respect to accuracy efficiency of MRI based Disease of Alzheimer detection by deep learning. In the study, the fine-tuned model compared with pre-trained models based on several performance matrices to show whether the proposed weighted-averaging is doing better for classifying and detecting AD classes. Furthermore, in the future, we can work on to ADNI or OASIS datasets and for the best result use Positron emitted tomography, diffusion tensor imaging neuroimages along with their combinations.